Yap Vooi Voon

Vision-based automated parking system

This paper describes an approach to overcome a situation of monitoring and managing a parking area using a vision based automated parking system. With the rapid increase of cars the need to find available parking space in the most efficient manner, to avoid traffic congestion in a parking area, is becoming a necessity in car park management. Current car park management is dependent on either human personnel keeping track of the available car park spaces or a sensor based system that monitors the availability of each car park space or the overall number of available car park spaces. In both situations, the information available was only the total number of car park spaces available and not the actual location available. In addition, the installation and maintenance cost of a sensor based system is dependent on the number of sensors used in a car park. This paper shows a vision based system that is able to detect and indicate the available parking spaces in a car park. The methods utilized to detect available car park spaces were based on coordinates to indicate the regions of interest and a car classifier. This paper shows that the initial work done here has an accuracy that ranges from 90% to 100% for a 4 space car park. The work done indicated that the application of a vision based car park management system would be able to detect and indicate the available car park spaces

Resource minimization in a real-time depth-map processing system on FPGA

Depth-map algorithm allows camera system to estimate depth. It is a computational intensive algorithm, but can be implemented with high speed on hardware due to the parallelism property. When depth-map algorithm is implemented on FPGA, resource consumption is one of the issues. The problem is normally resolved by modifying the algorithm, but the problem can also be solved by implementing new hardware architectures without modification of the depth-map algorithm. This work implemented five different processor architectures for the sum of absolute difference (SAD) depth-map algorithm on FPGA in real-time. Resource usage and performance of these architectures were compared. Memory contention and bandwidth constraints were resolved by using self-initiative memory controller, FIFOs and line buffers. Parallel processing was utilized to achieve high processing speed at low clock frequency. Memory-based line buffers were used instead of register-based line buffers to save 62.4% of logic elements (LEs) used. Usage of registers to replace repetitive subtractors saves 24.75% of LEs. The system achieves performance of 295 mega pixel disparity per second(MPDS) for the architecture with 640×480 pixels image, 3x3 pixels window size, 32 pixels disparity range and 30 frames per second. It achieves processing speed of 590 MPDS for the 64 pixels disparity range architecture. The disparity matching module works at the frequency of 10 MHz and produces one pixel of result every clock cycle.

Contactless heart rate monitor for multiple persons in a video

In this paper a contactless heart rate (HR) measurement system has been proposed. Algorithm has been developed to process video in real-time captured using a webcam. In the first step face and region of interests (cheeks) are detected. RGB color model is used for analysis; hence three color traces, R, G, B are obtained for the video. Fast Fourier transform is applied to the traces and peak frequency is detected after band pass filtering. The HR was calculated using the peak frequency. The results obtained from three channels of the RGB model are compared for their accuracy. It ahs been observed that the green channel gives better results. The algorithms were implemented; for multiple subjects in a frame and different illumination conditions. The algorithms were also tested; for different distances between the subject and the camera. The minimum percentage error of 3.1% is achieved; in the presence of movement and multiple persons at the relative distance of 70 cm.

Tracking of EEG activity using topographic maps

Electroencephalography (EEG) signal is generated by electrical activity of human brain. EEG topographic maps (topomap) give an idea of the brain activation. Brain Mapping may be used to relate the connectivity and functionality of the brain through imaging. Brain functional connectivity helps to find functionally integrated relationship between spatially separated brain regions. Brain connectivity can be measured by several methods. The classical methods calculate the coherence and correlation of the signal. Brain connectivity can also be measured by using nonlinear methods like mutual information, generalized synchronization and phase synchronization. In this paper, we have developed an algorithm to map neural connectivity in brain by using full search block matching motion estimation algorithm. We have examined the behavior of human brain throughout a specific activity using Oddball Paradigm. In the first step the EEG signal is converted into topomaps. The activation between consecutive frames is tracked using motion vectors. Vector median filtering is used to obtain a smooth motion field by removing unwanted noise. In each activation several paths between brain lobes have been tracked.

Comparison of motion field of EEG topo-maps for tracking brain activation

Human Brain can be divided into four different lobes: frontal lobe, parietal lobe, temporal lobe, and occipital lobe. Every lobe has its specific function. Electroencephalography (EEG) measures the electrical activity of human brain that is generated by synchronized activity of thousands of neurons. In this paper, we have developed an algorithm to track brain activity by using motion field generated by the Horn-Schunck (HS) optical flow algorithm. We have acquired EEG data from 20 subjects using oddball experiment to investigate the flow pattern of EEG signal (activity) across brain lobes during a specific activity. The EEG data is converted into EEG topographic maps (topo-map). The flow of EEG activity between consecutive topo-maps is estimated by using flow of brightness pattern in an image. A tracking algorithm is developed by tracking the motion vectors to examine the flow of EEG signal that depends on the overlapping of the motion field between current frame and next frames. The motion field is generated and the result of tracking is compared with full search (FS) block matching motion estimation algorithm. It has been observed that the motion field produced by HS method gives better results than the FS algorithm.

Performance Evaluation Metrics for Video Tracking

Abstract Various tracking methods have been developed to track objects with different degrees or levels of tracking ability. The ability or performance of each tracking method is dependent on the feature or data that is being used for tracking purpose. The ability of a tracking method can be measured by utilizing tracking metrics to give an indication of the tracking ability of an algorithm. This paper offers some insights into the issues and similarities of performance measurement reporting of video tracking algorithms and proposes a method in assessing the robustness of a video tracking algorithm. The proposed metric introduces another measure to measure the consistency of a tracking algorithm. The work presented in this paper shows that using only one metric to measure the tracking performance is inadequate. The proposed metric presented in this paper shows that the utilization of multiple metrics such as tracking success rate and tracking consistency or robustness would give a better indication of the tracking ability of a tracking algorithm used in video surveillance.

Image classification using wavelet coefficients for video surveillance

In this paper a new approach of wavelet coefficient features to identify and classify different images encountered by video surveillance systems is introduced. The main objective is to identify specific wavelet filters for achieving a better classification of images. The fundamental properties of wavelet coefficients for feature selection for image classification are investigated. Haar, Daubechies 5, Symlet 2, and Biorthogonal 2.2 wavelets have been used in this investigation. The results show that Haar wavelet provides promising results in object retrieval compared to Daubechies 5, Symlet 2, and Biorthogonal 2.2 wavelets.

Parametric tracking of multiple segmented regions

This paper proposes a tracking method based on parameters obtained from multiple blobs. The multiple blobs are derived or obtained from segmenting a single blob into multiple blobs or multiple regions that have the same color information where these regions generally remain the same as the target moves. The target being tracked is a person or human walking through the camera view field where the number of regions are dependent on the clothing worn and the general build of the person being tracked. This would indicate that the head, limbs and torso of a person would be segmented into regions of interest. The parameters used in tracking a multiple region or blob target are the vectors between the regions of interest and the mean values of the regions of interest. In this paper, the vector used is derived from the top most region to the lowest region of the multi region target. In addition, the difference between the mean values of the regions is used as a means of tracking in addition to the vector between the regions of interest. The results obtained showed that correct target tracking had consistently higher tracking rate compared to incorrect tracking. Correct tracking rates had higher than 0.9 Track Detection Rate (TDR) rates as compared to incorrect tracking that ranged from 0 to 0.6.